14494
clinical study

Performance Comparison of Two AI-Based Intracranial Hemorrhage Detection Tools on Head CT in a Real-World Clinical Setting

Materials & Methods
A retrospective six-week analysis (June 7–July 16, 2023) was performed at a single academic medical center to compare the diagnostic sensitivity and specificity of two FDA-cleared AI tools (Vendor A: Aidoc; Vendor B: Viz.ai). Consecutive noncontrast head CTs were included, excluding follow-up studies to maintain one-to-one patient–scan pairing. Discordant results between the two systems underwent adjudication by board-certified radiologists to establish reference truth. Vendor processing failures were documented.

Results
Among 1,192 CT scans, 147 cases were confirmed to have intracranial hemorrhage (ICH), corresponding to a prevalence of 12.3%. Vendor A demonstrated a sensitivity of 99.2%, specificity of 96.8%, positive predictive value (PPV) of 79.6%, and negative predictive value (NPV) of 99.9%. Vendor B demonstrated a sensitivity of 68.7%, specificity of 99.6%, PPV of 95.7% and NPV of 96.2%. A total of 87 discordant cases required adjudication, of which 43 were ultimately determined to be true positives. Vendor A failed to process 10 studies that were successfully analyzed by Vendor B.

Conclusions
Vendor A demonstrated markedly superior sensitivity while maintaining high specificity, highlighting clinically meaningful performance distinctions between FDA-cleared ICH detection tools. Understanding real-world variability in AI performance is essential for safe implementation, appropriate triage, and optimized stroke care workflows.

Disclaimer: This links to a third-party website, which may require credentials to review.

Explore the Latest AI Insights, Trends and Research